Improving end-use quality in hard winter wheat through glutenin allele combinations and genomic selection

Wheat (Triticum aestivum L.) has unique properties that allow for a variety of end products, such as pan bread, steamed bread, cookies, cakes, and tortillas. Most wheat-breeding programs focus on increasing yield and yield-related traits as primary objectives. However, end-use quality is also crucial as quality characteristics influence grain sale price and market success of a variety. Large-effect quantitative trait loci (QTL) have been identified for quality related traits. The Glu-1 loci encoding high molecular weight glutenin subunits (HMWGS) have a major effect on dough mixing properties. However, many quality traits are too complex to be controlled by only a small number of loci. These traits may benefit from genomic selection (GS), which utilizes all effective loci regardless of effect size. Genomic selection can accelerate genetic progress especially for traits that are costly or time consuming to phenotype, like quality-related traits. This research focused on the genetic improvement of end-use quality in hard winter wheat by targeting specific loci with known effects or by using all loci in a GS approach. The objectives of this study were to: i) evaluate agronomic and quality effects associated with different combinations of HMW-GS at the Glu-B1 and Glu-D1 loci among a set of near isogenic lines (NILs); ii) use a genome-wide association approach to identify QTL and develop predictive models for pre-harvest sprouting tolerance (PHST) and iii) assess GS models for milling and baking traits in hard winter wheat lines representative of west-central U.S. Great Plains germplasm. A set of NILs that varied for alleles at the Glu-B1 and Glu-D1 loci were evaluated for dough mixing properties, kernel characteristics, and agronomic effects. Results confirmed the Bx7OE + By8 HMW-GS (Glu-B1a1 allele) at Glu-B1 contributed to greater dough strength compared to the common Bx7 + By8 HMW-GS (Glu-B1b allele); however, the effect was not as significant as that conferred by Dx5 + Dy10 subunits (Glu-D1d allele). Near isogenic lines with the combination of both favorable alleles at Glu-B1 and Glu-D1 had the largest mixograph mixing time. However, a decrease in yield was observed for groups containing the Bx7OE + By8 subunits. These results suggest glutenin allele combinations are useful for improving bread-making characteristics in winter wheat but some combinations may be associated with negative effects on yield. Pre-harvest sprouting (PHS) is a major problem in wheat that results in decreased yield and quality. Genomic selection was evaluated as a potential breeding method for PHST given the complex inheritance and phenotyping difficulty of this trait. In this study, genotyping-by-sequencing (GBS) markers were used to identify QTL associated with PHST among a panel of hard red and white winter wheat lines. Genomic selection models were developed with the GBS data and phenotype data collected across seven growing seasons. The effect of including identified QTL and kernel color as fixed effects in the model was assessed, as kernel color has been generally associated with sprouting tolerance. Optimum marker number was also determined as accuracy can vary with different numbers of markers. Results showed model accuracy did not improve with kernel color information but weighting major QTL increased predictive performance. Optimum marker number was 4,000 with no improvement in accuracy above this threshold. Overall, model accuracies were promising and confirmed wheat breeding programs would benefit from incorporating GS models for PHST. Lastly, the accuracy of GS models for 11 end-use quality traits in a panel of hard red and white winter wheat breeding lines phenotyped across multiple years and locations was assessed. Trait heritability, marker number, and marker imputation method were evaluated for their effect on model accuracy. Traits measured included flour yield, single kernel characteristics, protein concentration, mixograph mixing time and tolerance, bake absorption, bake mixing time, crumb grain score, and loaf volume. Genotyping-by-sequencing marker data varied for marker density and imputation method used for missing data. Across traits, model accuracies ranged from 0.30 to 0.63 and trait heritability ranged from 0.03 to 0.61. Imputation method and marker density had little to no effect on model accuracy. Heritability appeared to have the greatest effect on accuracy as GS models for traits with higher heritability had higher accuracies. Additionally, GS models for moderate to high heritability traits performed better than expected when predicting a set of genotypes separate from the training panel. Results showed model accuracies for end-use quality traits were sufficient for increasing genetic gain in a wheat breeding program. In summary, genetic improvement in end-use quality can be made by utilizing both large effect and small effect loci in the wheat genome for such traits and will reduce phenotyping costs while increasing efficiency in a breeding program. In many winter wheat breeding programs, particularly those at higher latitudes, phenotypic quality evaluations from one season cannot be used for planting decisions of the next season due to the short turn-around time from harvest to planting. Genomic selection potentially solves this problem as selection decisions based on genotypic data can be implemented before the next season of planting. Thus, results from this study support the implementation of GS to reduce phenotyping costs and increase the rate of genetic gain for end-use quality in wheat.